Prognostic Value of the Neutrophil Percentage-to-Albumin Ratio for Mortality in ICU Patients With Myocardial Infarction: A Retrospective Cohort and Machine Learning Analysis

Zhantao Cao^1,2Zhonghui Lin²Xuejing Xu²Zhanglu Zhang²Xuanjing Chen²Jun Chen^2*Yunsu Wang^1,2*

• ¹Fujian University of Traditional Chinese Medicine, Fuzhou, Fujian Province, China
• ²Affiliated Xiamen Hospital of Traditional Chinese Medicine, Fujian University of Traditional Chinese Medicine, Xiamen, China

Background Although the neutrophil percentage-to-albumin ratio (NPAR) has shown prognostic value in multiple clinical conditions, its prognostic accuracy for myocardial infarction (MI) patients receiving intensive care has yet to be clearly defined. To our knowledge, this study is the first to comprehensively evaluate the prognostic role of NPAR in ICU-admitted MI patients, integrating both conventional Cox regression and machine learning approaches to address an existing gap between general MI cohorts and critically ill populations. Method Using data from the MIMIC-IV v3.1 database, we retrospectively included 1,759 ICU-admitted MI patients and calculated NPAR at admission. Primary and secondary outcomes were 30-day and 360-day all-cause mortality, respectively. Kaplan–Meier curves and log-rank tests compared survival across tertiles. Multivariate Cox models assessed associations, with restricted cubic splines evaluating nonlinearity. Machine learning models incorporating NPAR were developed to predict 30-day mortality, and model performance was assessed using the area under the receiver operating characteristic curve (AUC). Result The 30-day and 360-day all-cause mortality rates were 24% and 38%, respectively. Kaplan–Meier analysis revealed significantly lower survival probabilities in patients with higher NPAR levels. Adjusted Cox regression showed that those in the highest NPAR tertile had an increased risk of 30-day (HR: 2.03, 95% CI: 1.51–2.73, p < 0.001) and 360-day (HR: 1.81, 95% CI: 1.45–2.26, p < 0.001) mortality. Machine learning models incorporating NPAR achieved an AUC of up to 0.81 for predicting 30-day death. Conclusion The NPAR serves as an independent predictor of mortality at 30 and 360 days in MI patients admitted to the intensive care unit (ICU). When integrated into machine learning models, NPAR enhances predictive accuracy. These results indicate that NPAR serves as an independent predictor of short-and long-term mortality in ICU-admitted MI patients. Given its simplicity and accessibility from routine laboratory tests, NPAR can be feasibly incorporated into clinical decision-making and risk stratification protocols in critical care settings to facilitate individualized risk assessment and improve outcomes.